The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to handle and manage blockchain private keys for Sign-In With Ethereum (SIWE) authentication.
Evidence: Private keys are accepted via the NADMAIL_PRIVATE_KEY environment variable and stored locally in ~/.nadmail/private-key.enc.
Risk Mitigation: The skill uses AES-256-GCM encryption for local storage and sets restrictive file permissions (chmod 600), adhering to security best practices for local credential management.
[INDIRECT_PROMPT_INJECTION]: The inbox.js script retrieves and displays the body of incoming emails, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: Data is fetched from the external API endpoint https://api.nadmail.ai/api/inbox and processed in scripts/inbox.js.
Boundary markers: No explicit delimiters or instructions are used to separate the untrusted email content from the agent's internal logic.
Capability inventory: The skill possesses the ability to send emails and trigger financial transactions ('emo-buy') via scripts/send.js, which could be exploited if the agent is compromised by malicious email content.
Sanitization: No sanitization or filtering of the email body or snippets is performed before outputting to the console.
[COMMAND_EXECUTION]: The skill consists of multiple Node.js scripts intended to be executed by the agent to perform registration, sending, and reading operations.
Evidence: Scripts include setup.js, register.js, send.js, and inbox.js.
Safety Controls: The skill includes an 'emo-buy' safety mechanism in send.js with a default daily spending cap of 0.5 MON and a requirement for user confirmation unless the --yes flag is explicitly provided.

NadMail